The invention relates to an apparatus and method for controlling a welding process. More specifically, the invention relates to apparatus and methods for viewing a welding operation, determining the width of a weld pool, for controlling an arc welding process (including control of weld penetration) and for locating a weld line. The invention also relates to certain data handling techniques which are useful in the instant welding control process but which may also be useful in other applications.
The term "weld line" is used herein to mean the line between the adjacent edges of two metal portions in a weld joint preparation. Thus, the weld line is the line which the welding apparatus should follow during the welding operation. Those skilled in the art will appreciate that, in some types of weld joint preparations, there will be a gap of finite width between the metal portions being welded and thus the weld line will in fact be a narrow strip rather than a line in the strict mathematical sense of that term. The term "weld joint preparation" is used herein to denote the two metal portions lying adjacent one another in a position ready for welding.
The process of arc welding is one of the most widely used manufacturing processes in the world. Great efforts have been made in recent years to automate the welding process in order to allow the use of robot welders which are capable of producing very uniform, high quality welds. Such automation has the potential both to increase productivity by increasing the speed and accuracy of the welding process in routine welding applications, and to increase the quality of the welds thus produced, thereby allowing automated welding to be utilized in high quality control industries such as the nuclear power and pressure vessel manufacturing industries.
In order to develop automatic control of the arc welding process, numerous attempts have been made to determine which are the relevant parameters in the welding process. It has truly been said that only the welding operator truly knows the quality of the weld and what is happening in the welding process because only he has hitherto been able to see precisely the quality of the weld and the physical parameters of weld pool size and arc length. Also, only the operator has been in a position to accurately track weld lines which are often irregular in shape.
In recent years, use of television based direct monitoring has increased. All prior art television monitoring methods have positioned the camera so as to provide an oblique view of the weld similar to that seen by an operator. Unfortunately, such an oblique view inherently includes the electric arc, which is very bright and thus tends to "wash out" the entire television picture unless appropriate filters are utilized. Therefore, these prior art methods have required filtering the light reaching the camera so that only light within a narrow band, or of several discrete wavelengths, reaches the camera, rather than allowing the camera to receive light of the broad range of wavelengths emitted by the arc.
Additional problems in obliquely viewing the weld area with a television monitoring camera are that unforeseen obstructions of the view seen by the camera can arise in real life welding situations can arise. Such obstructions can be caused by the weld line geometry and by constraints on the placement of the camera. Also, distortion of the image of the weld pool is caused by the oblique viewing position due to parallax effects.
Prior art weld pool monitoring techniques such as those already discussed have been used for a variety of purposes. One such purpose is the monitoring of weld pool width. Prior methods of analysis of video data received from viewing the weld pool area with a television camera have made the assumption that the bright areas represent the weld pool and that once the light intensity has decreased to a certain value, then the edge of the weld pool has been reached. Such methods then employ a "go/no-go" type of binary logic system to establish the weld pool width. Unfortunately, there are oscillations in the weld pool caused by fluctuations in the welding current, and also caused by the motion of the electrode along the weld line. Such oscillatory motions cause the area of brightness to appear larger than it actually is. Attempts to mitigate the data error caused by the oscillatory motions have been effected simply by correcting the data rather than actual measurement and evaluation of the true weld pool edges.
A further purpose to be served by the weld pool monitoring techniques already discussed is to control penetration of the weld i.e. the depth of the melted zone of the weld. It is known that there is a relationship between the width of the weld pool and the penetration achieved by the welding process and thus it has been thought desirable to be able to measure and control weld pool width to effect correspondingly precise control of the penetration being achieved. In order to provide a weld of adequate strength, weld penetration must be an appropriate percentage of metal thickness to produce proper adhesion of the two metal portions forming the weld joint preparation when the molten metal of the weld pool has resolidified. Full penetration welds are welds in which the penetration extends all the way through the metal portions. More commonly, only partial weld penetration is necessary. However, hitherto the only methods of measuring partial weld penetration have been intelligent guesswork by an experienced operator or destructive testing techniques following welding. Such destructive testing techniques involve making a sample weld, cutting a cross-section therethrough and actually measuring the weld penetration. Although recently attempts have been made to use non-destructive testing techniques, such as ultrasonic and radiographic analysis, to measure weld penetration, most methods of non-destructive testing yield inconclusive results when applied to a partial penetration weld, and thus such non-destructive testing techniques are usually only used for analysis of full penetration welds. Because of the difficulty of estimating the penetration of partial penetration welds, engineers have mandated full penetration welds when they are concerned about the structural integrity of the final welded product, even though, if proper control of weld penetration could be achieved, only partial penetration welds of specified penetration could be used, thereby increasing the speed and efficiency of the welding process while still allowing close design control. There is thus a need for a method of measuring weld penetration during welding.
A further problem with prior art methods for controlling welding processes is that, if they incorporate computerized analysis of video data, they tend to make excessive demands upon the capacity and/or speed of memory devices used to handle the video data to be analyzed. Storage of every complete frame of video data received from a typical video camera presents extreme difficulty both as to the capacity and speed of the memory required. For example, storing a typical video frame, using an eight-bit graytone scale, requires approximately 4 million bits of information per frame, and such frames may be arriving at, typically, 25 Hz. One technique for reducing data storage requirements is only to store the idea of each frame which is of interest, but if this is to be done it is necessary to allow ready variation of the "window" selected for data storage if the analysis method is to be sufficiently flexible. Also, there is the problem that the rate at which video data from individual pixels of the image to be analyzed is received by memory may be quicker than the memory can handle such data, and to overcome this problem it is desirable to provide a method whereby a stream of data being fed to a memory at a higher rate can be split among a plurality of relatively slow memories each individually incapable of handling and storing the incoming stream of data.
Finally, in an automated welding process it is necessary to ensure that the welding apparatus accurately tracks the weld line. Attempts have been made to track weld lines by means of sensors which look at the line ahead of the welding torch. Such sensors attempt to locate the position of the weld line relative to the torch and adjust the movement of the welding torch so that it tracks the weld line. Such methods have utilized both direct contact type sensors which are "dragged" along the weld line ahead of the welding torch and non-contact sensors such as infra-red detectors.
In prior art systems for tracking weld lines, in order to avoid damage to the sensor and also to keep the sensor from being obscured by the light from the arc area, it has been necessary to sense the preparation some finite distance ahead of the weld torch. This immediately produces the requirement of having some delay in system response so that in fact the system responds to seam tracking changes at the time when the welding head is over the area of change, not just as they are sensed.
Prior art weld line tracking devices are of two types. The first type senses voltage and current variations in the arc when various surface features of the base metal being welded are encountered. One such system oscillates the arc back and forth across the weld line noting the voltage changes due to arc length, the arc length becoming longer as it crosses the weld line. Various schemes have been proposed to measure this voltage oscillation; both magnetic and mechanical motion devices have previously been utilized. The second type identifies the edges of the metal positions on either side of the weld line by analysis of light from the arc reflected off these edges. A feedback system is then provided to respond to the reflected light to produce the desired effect of following the weld line. Of course, since the light source is the arc itself, the welding process is already occurring at the weld line and the system simply keeps the welding torch over the weld line. Since this type of system tracks the weld line by viewing the line ahead of the welding torch, means must be provided to delay implementation of any necessary changes in movement of the welding torch until the torch reaches the appropriate part of the weld line. Also, any process control data produced by such apparatus are subject to error due to parallax from the oblique camera position, as already described, and the apparatus suffers from the problem that the far side of the weld pool is hidden by the arc itself.
Additionally, when prior art methods for tracking a weld line indicate a change in direction of the weld line, the entire welding assembly including the tracking device and torch must be reoriented so that the tracking device leads the welding torch at all times. Such a system inherently complicates the weld tracking control system since the torch must not only be translated to follow the weld line, but it must also be rotated so that the tracking device leads the torch head.
Therefore, there is a need for an improved weld line tracking apparatus and method that reduces the delay time between receiving a response and actually moving the torch head. Additionally there is a need for a system which will allow simple rectilinear translation of the torch head without the necessity of also rotating the torch head and tracking device.
This invention seeks to provide apparatus and methods for overcoming the aforementioned problems in controlling welding processes.